Primitive Chain Network Simulation of Elongational Flows of

Dec 1, 2011 - mechanisms helped us to understand entangled polymer dynamics ... tional flow behavior was examined with the primitive chain network (PC...
0 downloads 0 Views 2MB Size
ARTICLE pubs.acs.org/Macromolecules

Primitive Chain Network Simulation of Elongational Flows of Entangled Linear Chains: Role of Finite Chain Extensibility Takatoshi Yaoita,†,‡,^ Takeharu Isaki,‡ Yuichi Masubuchi,† Hiroshi Watanabe,†,* Giovanni Ianniruberto,§ and Giuseppe Marrucci§ †

Institute for Chemical Research, Kyoto University, Kyoto 611-0011, Japan Material Science Laboratory, Mitsui Chemicals Inc., Chiba 299-0265, Japan § Dipartimento di Ingegneria Chimica, Universita “Federico II”, P. le Tecchio 80, 80125 Napoli, Italy ‡

ABSTRACT: For entangled linear monodisperse polymers, uniaxial elongational flow behavior was examined with the primitive chain network (PCN) simulation, which was originally formulated for a network of Gaussian chains bound by sliplinks but was modified in this study to properly take into account the finite extensibility of actual chains. On an increase of the elongational rate ε· from the terminal relaxation frequency 1/τd (at equilibrium) to the Rouse relaxation frequency 1/τR, the original and modified simulations gave an indistinguishable steady state elongational viscosity ηE that almost scaled as ε·1/2. On a further increase of ε· > 1/τR, ηE obtained from the original PCN simulation diverged to infinity (as noted also for the unentangled Rouse chains). In contrast, ηE deduced from the modified simulation increased but did not diverge with increasing ε· > 1/τR (similarly to the behavior of FENE dumbbells). This feature of the modified PCN simulation, i.e., hardening to a finite (nondiverging) level, mimicked the ηE data of entangled semidilute solutions, which naturally reflected the finite extensibility of actual chains. Analysis of the simulation results suggested that the power law behavior (ηE ∼ ε·1/2) at 1/τd < ε· < 1/τR is related to reduction of the entanglement density and the corresponding reduction of chain tension, while the hardening (upturn of ηE at ε· > 1/τR) results from stretch of the chain (eventually approaching full stretch), thus shedding light on the behavior of semidilute solutions. Nevertheless, the modified simulation did not describe the behavior of entangled melts, i.e., ηE ∼ ε·1/2 even at ε· > 1/τR. A factor missing in the modified simulation is discussed in an attempt to elucidate, from a molecular point of view, the difference between entangled solutions and melts.

1. INTRODUCTION For entangled polymers, extensive studies have been made experimentally and theoretically to elucidate, from a molecular viewpoint, the role of entanglements. In particular, the tube model refined so as to include a variety of molecular relaxation mechanisms helped us to understand entangled polymer dynamics to a considerable depth.14 Nevertheless, several important experimental observations have not been fully addressed within the context of the refined tube model and remain as subjects of theoretical challenge. One of the current challenges, a major one, relates to the uniaxial elongational flow behavior of entangled monodisperse linear chains. Figure 1 schematically illustrates the strain rate (ε·) dependence of the steady state elongational viscosity (ηE) of monodisperse linear polystyrene (PS) reported in the extensive work by Sridhar’s group (for entangled semidilute solutions)57 and by Hassager’s group (for entangled melts).8,9 With increasing ε· from the reciprocal of the terminal relaxation time τd at equilibrium (often interpreted as the reptation time) to the reciprocal of the Rouse relaxation time τR of the whole chain, ηE of both solutions and melts decreases, essentially scaling as ηE  ε·1/2. However, on a further increase of ε· > 1/τR, ηE of the entangled solutions increases (without diverging) to a level even above the linear-regime viscosity ηE,0 (= 3η0), whereas ηE of the r 2011 American Chemical Society

entangled PS melts does not show this hardening behavior but keeps decreasing roughly as ηE  ε·1/2. Here we rely on the experimental data mentioned above, although PS melt data from a different laboratory10 also confirm Hassager’s data. As regards possible differences between solution and melt experiments, we recall that both data sets were taken by the filament stretching rheometer with similar feedback systems to achieve constant elongational rates. Some minor differences may exist in sample adhesion to the end-plates and in thermal equilibration, but such aspects certainly cannot explain the marked differences in rheological behavior schematically reported in Figure 1. Theoretical attempts have been made for explaining this difference of the elongational flow behavior of entangled solutions and melts. For example, the tube model refined by Mead, Larson, and Doi (MLD)11 accounting for the nonequilibrium mechanism of convective constraint release (CCR), 12 and the resulting dynamic tube dilation (DTD),13 can mimic the ηE data of the entangled solutions, as reported by Sridhar and co-workers.57 However, the ηE data of the entangled melts do not appear to be consistently explained by the MLD model Received: September 26, 2011 Revised: November 13, 2011 Published: December 01, 2011 9675

dx.doi.org/10.1021/ma202166y | Macromolecules 2011, 44, 9675–9682

Macromolecules

ARTICLE

the PCN code, with the expectation that this modification would allow the simulation to predict the hardening but nondiverging ηE thus mimicking the data of the entangled solutions schematically shown in Figure 1. We do not expect, however, that the modified simulation successfully describes the melt data, and we will discuss possible factors distinguishing melts from solutions, postponing a detailed analysis of such factors to a subsequent paper.

Figure 1. Schematic illustration of uniaxial steady-state elongational viscosity ηE of entangled melts and semidilute solutions of monodisperse linear chains. ηE,0 is the steady-state viscosity in the linear range.

even after a reasonable change of the model parameters. Wagner and co-workers9 incorporated the concepts of interchain pressure (ICP) due to tube confinement and of strain-dependent tube diameter14,15 in their molecular stress function model and showed that the ηE data of the melts can be described by this model. Nevertheless, the tube diameter relaxation time required for description of the data has not been fully explained within the context of the model, and remains as an adjustable parameter. In addition, incorporation of both CCR and ICP effects in the tube model could result in underestimation of the ηE data of the melts,16 despite the fact that CCR is an unavoidable ingredient of the tube model under flow. It should be noted that the extensional behavior of dilute solutions has been extensively studied as well.1722 For example, Gupta et al.21 have reported that some dilute and semidilute (unentangled) solutions also show the ηE  ε·1/2 behavior in the range of ε· > 10/τR. This behavior has been explained by the chain finite extensibility20 and by the friction anisotropy.18,21 However, it is well-known that the molecular mechanisms relevant for entangled solutions and melts are very much different from those prevailing in dilute solutions.1,4 In view of the above recalled uncertainties, this paper analyzes the many-chain dynamics under elongational flow with the aid of the primitive chain network (PCN) simulation23 (i.e., by replacing the tube confinement with binary sliplinks). The PCN simulation successfully described the linear viscoelastic and nonlinear shear thinning behavior2327 as well as the nonlinear damping behavior28 of entangled linear chains. In the simulations with the original PCN code,2326,28 the entangled system was modeled as a network of Gaussian chains bound by sliplinks, and these chains were allowed to move/relax through reptation/ contour length Rouse fluctuation through the sliplinks,1 as well as through CCR12 and thermal CR29 induced by removal and creation of sliplinks. In this study on uniaxial elongational flows, we used a modified PCN code that accounts for finite extensibility, as also done by Shaqfeh and co-workers in their multichain sliplink model when examining the nonlinear behavior under shear30 and planar31 extensional flows. More specifically, in this study we first conducted the PCN simulation with the original Gaussian code under uniaxial elongational flow conditions in the expectation that the calculated ηE decreases essentially as ε·1/2 for 1/τd 1/τR because of the infinite extensibility of Gaussian chains. Since actual chains have an extension limit,30,31 we then incorporated finite extensibility in

2. MODEL In the PCN simulation,2328 sliplinks are distributed along the polymer chains to divide them into subchains having a molecular weight comparable to the entanglement molecular weight. Each sliplink bonds two chains, consistently with the binary assumption for entanglements. The sliplink is allowed to move in space according to a balance among drag force, subchain tensions, osmotic force, and random thermal force. The polymer chain slides through the sliplinks (i.e., reptates) to fulfill the force balance along the chain. If the end of the chain moves past a sliplink, that sliplink is removed. On the contrary, if the monomers (taken to be the Kuhn segments) reptate, by a certain amount, beyond the sliplink located at the chain end, a new sliplink is created by hooking one of the surrounding chains. In summary, dynamics of the system is ruled by: 3D motion of the sliplinks, monomer reallocation between adjacent subchains via chain sliding through the sliplinks, and creation/removal of sliplinks at chain ends. The above dynamics is described by the stochastic equations summarized below, with the finite extensibility of the chains (not considered in the original Gaussian PCN code) affecting the equations for sliplink position and monomer number per subchain through the expression for the elastic force. 2.1. 3D Motion of Sliplinks. The sliplink position, R, obeys the time-evolution equation representing the force balance mentioned above:23 ζðR_  jRÞ ¼ FB þ Ff þ

4

∑i Fi

ð1Þ

Here, R_ is the time derivative of R, and j the velocity gradient tensor. FB is the thermal random force, Ff the osmotic force due to concentration fluctuation of the sliplinks, and Fi the elastic force in the i-th subchain of the 4 subchains acting upon the sliplink of our focus. The friction coefficient ζ of the sliplink is assumed to be given by ζ¼4

n0 ζ 2 m

ð2Þ

where n0 is the average number of monomers (Kuhn segments) in the subchain at equilibrium, and ζm is the monomeric friction coefficient. The factor of 4 accounts for the number of subchains converging to the sliplink, while the factor 1/2 reflects the fact that the subchain friction is split into the two sliplinks confining the subchain. Among the forces appearing in eq 1, FB is modeled as a white noise having zero mean and the second-order moment given by,23 ÆFB ðtÞFB ðt 0 Þæ ¼ 2kTζδðt  t 0 ÞI

ðI ¼ unit tensorÞ

ð3Þ

with kT being the thermal energy. Ff is related to the chemical potential μ as Ff = rμ. The free energy A, giving 9676

dx.doi.org/10.1021/ma202166y |Macromolecules 2011, 44, 9675–9682

Macromolecules

ARTICLE

μ as a derivative, is of the simple form: 8 !2 > > Æϕæ Æϕæ A¼ > > : 0 for ϕ e Æϕæ

The elastic tension Fi appearing in eq 8 is related to ri and to the factor f defined by eqs 6 and 7 (without/with the finite extensibility) through the scalar equivalent of eq 5: ð4Þ

In eq 4, ϕ is the local subchain density (number of subchains in the entanglement volume a3 =(b(n0)1/2)3, with b = Kuhn length), Æϕæ is the average of ϕ in the whole system, and ε is a numerical factor specifying the magnitude of Ff (ε = 0.5, following a previous study25). Finally, Fi is the elastic force of the i-th subchain acting on the sliplink27 Fi ¼ f

3kT ri ni b2

ð5Þ

with ri being the end-to-end vector and ni the monomer number of the i-th subchain. If the subchain is treated as the infinitely extensible Gaussian chain (as in the original PCN code23), the factor f appearing in eq 5 is given by f ¼1

ð6Þ

On the other hand, for the actual subchains with the finite extensibility, f can be expressed within the FENE-P approximation as33 f ¼

1 1  Æðri =ri, max Þ2 æi

with ri ¼ jri j

ð7Þ

Here, ri,max (= bni) is the fully stretched i-th subchain length. The use of FENE-P rather than FENE eliminates numerical difficulties due to overstretched segments artificially generated by Brownian kicks and/or conformational rearrangements upon creation of sliplinks. Obviously, the difference between eqs 6 and 7 has a significant effect on the PCN simulation results only when the subchains are considerably stretched. In relation to eq 7, we should point out that in a previous work27 we used the PCN model with FENE rather than FENE-P. The use of FENE requires several changes in the constraint renewal algorithm (to be explained in section 2.3) and such changes induce differences in the basic parameters (units of length and time27) as well as in chain statistics.34 Since we have confirmed that, in spite of those differences, the rheological responses predicted by the PCNFENE and PCN-FENE-P codes are similar to one another, for simplicity we use here the FENE-P approximation. 2.2. 1D monomer reallocation through sliplinks. Subchain monomers are redistributed as a result of differences of tensions of adjacent subchains, of chemical potential gradients along the chain, and of one-dimensional Gaussian random force. The rate of sliding of monomers (_n) from the (i1)-th to the i-th subchain through the sliplink is described by: n_ ζs ¼ ðFi  Fi1 Þ þ fB þ ff F

ð8Þ

In eq 8, F is the linear density of monomers averaged over the (i1)-th and the i-th subchains, and ζs is the friction coefficient for the sliding motion, simply computed by taking half the (equilibrium) monomers in each subchain: n0 ð9Þ ζs ¼ 2 ζm ¼ n 0 ζm 2

Fi ¼ f

3kT ri ni b2

ð10Þ

The other force appearing in eq 8, ff, is the scalar equivalent of Ff, i.e., the field force along the chain due to fluctuations in subchain density.23 Finally, fB is the 1D Brownian force characterized by: ÆfB ðtÞfB ðt 0 Þæ ¼ 2kTζs δðt  t 0 Þ

ð11Þ

Recently Uneyama and Masubuchi35 have proposed a modification of the monomer sliding equation, eq 8, for better thermodynamic consistency (use of chemical potential differences instead of tension balance). However, extension of this modification to fast flows appears nontrivial, and therefore we here stick for simplicity to the standard PCN code to focus on the effect of finite extensibility. 2.3. Constraint Renewal. Each chain creates and destroys sliplinks at its ends, thus renewing the topology of the entangled network (constraint renewal). In the simulation, creation and/or destruction of sliplinks results from monitoring the monomer number in the end subchains. The end subchain creates a new entanglement when its monomer number exceeds 1.5n0, while it removes its entanglement when the monomer number falls below 0.5n0.24 Although this algorithm is clearly somewhat arbitrary, and its relationship to thermodynamics (detailed balance) remains unclear, it recommends itself for its simplicity and for its good performance in reproducing both linear2426,34 and nonlinear26,28 rheology of linear polymers. 2.4. Simulations. In the simulation, the Langevin equations, eqs 1 and 8, were solved numerically to follow the time evolution of {Ri}, {ri}, and {ni} and to calculate the stress tensor σ, σ ¼

1 V

∑i Fi ri  pI ¼ V

1 3kT f b2

∑i

ri ri  pI ni

ð12Þ

Here, the sum extends over all subchains in the volume V, f is the FENE factor defined by eqs 6 and 7, and p is the isotropic pressure (that balances the chain stress components perpendicular to the elongational direction). In the actual simulation, eqs 1 and 8 were nondimensionalized with the aid of the units of length, thermal energy, and time, a = b(n0)1/2, kT, and τ0 = n02b2ζm/3kT, respectively. (Consequently, σ was obtained in the nondimensionalized form.) The simulation conditions, such as the number density of subchains per unit volume and the periodic boundary condition, are the same as in our previous work.23 A simulation box with dimensions of 8  8  8 was used for calculations of linear viscoelasticity. For uniaxial deformation, a flat simulation box 4  104  104, with 4 in the stretching direction, was used as the starting box, and the simulation was stopped either when a steady flow was achieved or when the box elongational ratio Λ reached the maximum value Λmax = 676 (i.e., when the deformed box became 2704  4  4) corresponding to an Hencky strain (εH = lnΛ) of ca. 6.5. The simulation box size did not affect the results, as explained in Appendix A. Equilibration of the starting box was achieved by simulation under quiescent conditions for a sufficiently long time (longer than the longest relaxation time of the system). For a chain with molecular weight M, the number of subchains per chain is given by Z0 = M/M0, with M0 being the average molecular weight of the 9677

dx.doi.org/10.1021/ma202166y |Macromolecules 2011, 44, 9675–9682

Macromolecules

ARTICLE

subchains. The average monomer number per subchain, n0, was calculated from the Kuhn molar mass MKuhn as n0 = M0/MKuhn.23 The length of the fully stretched subchain is given by ri,max = bni and its nondimensional expression is related to n0 as ~r i,max = bni/b(n0)1/2 = ni/(n0)1/2. The parameters required for the nondimensional simulation are Z0 and n0. From the modulus G0 = FRT/M0 we obtain M0, hence Z0 from M. Finally n0 is obtained from MKuhn. Comparison of the nondimensional simulation results with data also requires the unit of timeτ0. The parameters G0 and τ0 were unequivocally evaluated by fitting the linear viscoelastic data with the PCN simulation (under no strain), as explained in Appendix B.

3. RESULTS AND DISCUSSION 3.1. Overview of Data and Simulation Results. In Figure 2, symbols show the data of the viscosity growth function η+E(t,ε·) reported for a polystyrene (PS) solution (M = 3.9  106, c = 10 wt %)6 and a PS melt (M = 3.9  105).8 The curves indicate

the PCN simulation results with (solid curves) and without (dashed curves) FENE. The simulation parameters, determined from the fitting of the linear viscoelastic data (cf. Appendix B), are listed in Table 1. Figures 3 and 4, respectively, show the corresponding steady state elongational viscosity, ηE(ε·) = η+E(∞,ε·), of the solution and melt. Both the measured (symbols) and simulated (curves) viscosities are plotted against the Weissenberg number defined with respect to the Rouse relaxation time τR: WiR ¼ τR ε_ with

τR ¼

Z0 2 τ0 2π2

ð13Þ

(The Rouse time τR calculated from eq 1327 is also reported in Table 1.) As noted in Figure 2, the η+E(t,ε·) data of both the solution and the melt fall above the linear η+E(t,0) data shown in Figure 2 with

Figure 3. Steady-state uniaxial viscosity ηE for monodisperse linear PS solutions with M = 3.9  106 and c = 7.35 wt % (left panel) and 10 wt % (right panel) at 21C. The ηE data are normalized by the zero-strain-rate viscosity, ηE,0, and plotted against τR-based Weissenberg number WiR. The data obtained from ref. 5 are shown with the circles. Simulation results with and without FENE are indicated with the solid and dashed curves. Arrows indicate divergence for the Gaussian chain at WiR . 1 deduced from the simulation without FENE.

· Figure 2. Transient uniaxial viscosity η+E(t,ε) for (a) solutions and (b) melts of monodisperse linear polystyrene (PS). Experimental data for the solution (symbols) are obtained from ref. 6: M = 3.9  106, c = · 10 wt %, T = 21 C, and ε = 0.1, 3.5, 7.3, and 11.7 s1. Experimental data for the melt (symbols) are obtained from ref. 8: M = 3.9  105, T = · 130 C, and ε = 0.0003, 0.001, 0.003, 0.01, and 0.03 s1. Simulation results with and without FENE are indicated with the solid and dashed curves, respectively. Thin dotted curve indicates η+E(t) in the linear viscoelastic regime.

Figure 4. Steady-state uniaxial viscosityηE for monodisperse linear PS melts with M = 2.0  105 (PS200 K, left panel) and M = 3.9  105 (PS390 K; right panel) at 130C. The ηE data are normalized by the zero-strain-rate viscosity, ηE,0, and plotted against τR-based Weissenberg number WiR. Experimental data obtained from ref. 8 are indicated with the circles. Simulation results with and without FENE are indicated with the solid and dashed curves. Arrows indicate divergence for the Gaussian chain at WiR. 1 deduced from the simulation without FENE.

Table 1. Parameters of PS Samples (Solutions5,6 and Melts8) Used in the Simulations

a

sample

Mw

Z0

3.9 M 7.35 wt % sol.

3.9  10

3.9 M 10 wt % sol.

3.9  10

PS200 K PS390 K

2.0  10 3.9  105

6 6 5

18 33 18 35

τ0 [s]

n0

G0 [Pa]

300

8.1  10

2

160

2.0  10

3

15 15

2.9  10 2.9  105 5

τR [s] 2

2.7  101

3

4.6  101

a

9.1  101 (1.1  102)a 3.5  102 (4.3  102)a

1.7  10 8.3  10

5.6 (7.0) 5.6 (7.0)a

FENE for melts. 9678

dx.doi.org/10.1021/ma202166y |Macromolecules 2011, 44, 9675–9682

Macromolecules the thin dotted curves (=3η+(t) obtained from the linear viscoelastic response reported in Appendix B) at short to intermediate times. This transient hardening behavior is commonly observed for solutions and melts at WiR > 1 (ε· > 1/τR). However, in the steady extensional state, the solution and the melt of Figure 2 exhibit a significant difference: ηE (ε·) of the solution (Figure 3) exhibits an upturn (steady state hardening) and increases with increasing WiR in the range WiR > 1, while ηE (ε·) of the melt (Figure 4) keeps decreasing with increasing WiR even at such high WiR (roughly as WiR1/2, i.e., as ε·1/2). Figure 2a demonstrates that the transient strain-hardening behavior of the solution is described by the simulation either with or without the finite chain extensibility (with eq 7 or eq 6). However, the behavior of the steady state ηE (ε·) data (thinning followed by the hardening without divergence) can be mimicked only when the finite extensibility is incorporated in the simulation (cf. Figure 3). This conclusion is fully confirmed by the subchain stretch obtained from the simulation, reported in Figure 5a as the reduced stretch ratio ~λ = λ/λmax = Æri/ri,maxæi. In the early stage in the transient zone, the subchains in the solution are far from the fully stretched limit ~ is well below unity) so that the simulated η+E(t,ε·) is not (λ significantly affected by the finite extensibility. However, in the steady flow state at large WiR (.1), the simulation without the finite extensibility allows the subchains to be infinitely stretched to give the diverging ηE (ε·). (This diverging feature is well-known for the Gaussian chain modeled as a Rouse chain.) In contrast, the simulation accounting for the finite extensibility of the actual chains can reproduce semiquantitatively both thinning and nondiverging hardening of the ηE(ε·) data (cf. Figures 2a and 3). Considering this success of the simulation with the finite extensibility, the simulated crossover from the thinning to the hardening zone is analyzed in the next section to discuss the similar crossover observed for the solutions. (Note that the PCN simulation with the original code was made also for the nonlinear stress relaxation under step strain28 but the chain stretch under the conditions examined was well below the fully stretched limit. Thus, the results of those simulations are not affected by the FENE correction.) Differently from solutions, the transient strain hardening of the melt is followed by the thinning of ηE (ε·) ε·1/2 even at WiR . 1 (cf. Figures 2b and 4). This thinning at large WiR cannot be described by the simulation either with or without the finite extensibility, although the thinning at WiR of order unity appears to be better mimicked if the finite extensibility is incorporated in the simulation. Indeed, the simulations do not distinguish between melts and solutions, as also shown by the behavior of ~λ in Figure 5 where the melt and the solution are very similar. It so appears that some essential physical factor distinguishing entangled solutions and melts is missing in the current PCN simulation. Possible physical factors responsible for this difference are discussed in the last section. As mentioned in the Introduction, finite extensibility had already been implemented in earlier sliplink models, and used in extensional flow simulations.27,31,32 In particular, the singlechain sliplink model of Schieber et al.32 reproduces the ηE  ε·1/2 behavior in the range of ε· > 1/τR for melts (though departing considerably from data in the transient behavior), yet a behavior similar to melts is predicted for solutions as well, thus strongly underpredicting the solution elongational thickening. Kushwaha and Shaqfeh31 performed

ARTICLE

Figure 5. Simulation results for the transient stretch ratio λ normalized to λmax (fully stretched subchain) corresponding to the viscosity results shown in Figure 2. (a) PS solution with M = 3.9  106 and c = · 10 wt % at ε = 0.1, 3.5, 7.3, and 11.7 s1 from right to left. (b) PS melt · with M = 3.9  105 (PS390 K) at ε = 0.0003, 0.001, 0.003, 0.01, and 1 0.03 s from right to left. Results with and without FENE are shown with the solid and dashed curves, respectively. Arrows indicate divergence for the Gaussian chain at WiR . 1 deduced from the simulation without FENE.

multichain sliplink simulations under planar elongational flow and compared their results with melt data,8 observing power laws similar to ηE  ε·1/2 in the nonlinear range, and an upturn in the viscosity only at ε· . 1/τR.. Although the authors31 do not explicitly compare their predictions with entangled solution data, a “delayed” upturn is certainly in contrast with solution data in uniaxial elongational flow. To our knowledge, the present paper demonstrates for the first time a quantitative agreement of sliplink simulations with uniaxial extensional viscosity data of entangled solutions, showing at the same time the limitation of the current PCN code for dealing with entangled melts in uniaxial elongational flow. 3.2. Origin of Thinning Followed by Nondiverging Hardening of Solutions. Since the PCN simulation with FENE mimicked the elongational behavior of entangled solutions, analysis of the simulation results allows us to examine the factors responsible for the thinning-thickening crossover of the solution. We start from the expression of the stress tensor σ (cf. eq 12 with eq 7 for FENE), where for convenience of this analysis we adopt a decoupling approximation, and express the reduced tensile stress σErd as σ rd E 9679

  1 Z 2  ðσxx  σyy Þ = 3f λ S νkTZ0 Z0

ð14Þ

dx.doi.org/10.1021/ma202166y |Macromolecules 2011, 44, 9675–9682

Macromolecules

ARTICLE

Figure 6. Average stretch ratio λ, orientation anisotropy S, FENE factor f, and reduced entanglement density Z/Z0 obtained from the PCNFENE simulation for the entangled solutions (dashed and solid curves for c = 7.35 and 10 wt %, respectively) in the steady flow state as a function of WiR. For comparison, the simulated viscosity ηE is also shown (topmost panel).

where ν is the (constant) number of chains per unit volume, and Z/Z0 as well as  2 1=2   ðri ri Þxx  ðri ri Þyy ri , S ¼ , λ¼ ni b2 i jri j2 i f ¼

1 1  Æðri =rimax Þ2 æi

ð15Þ

are quantities variable with the stretching rate. In Figure 6, the average stretch ratio λ, the orientation anisotropy S, the FENE factor f, and the reduced entanglement density Z/Z0, all obtained from the PCN-FENE simulation for the entangled solutions in the steady flow state, are plotted against WiR. For comparison, the simulated viscosity ηE is also shown. Figure 6 shows that, in the range WiR < 1, both λ and S increase whereas Z/Z0 decreases with increasing WiR, and f remains close to unity. Now, since the viscosity ηE is given by ηE σE/ε· = σEτR/WiR, the decrease of Z/Z0 significantly adds to the decreasing factor 1/ WiR in overwhelming the increase of S (enhancement of orientation) and λ (enhancement of stretch) to give the thinning behavior of ηE. Thus, the thinning behavior is attributable, to a significant extent, to the CCR mechanism that results in the decrease of the entanglement density Z/Z0 and in the corresponding reduction of the chain tension, the latter being given by the product of three factors, i.e., λfZ/Z0. In contrast, in the range of WiR > 1, λ and f increase more rapidly with increasing WiR while the decrease of Z/Z0 remains similar to that in the range WiR < 1. Thus, the hardening of ηE seen at WiR > 1 is governed by enhancement of the chain tension due to chain stretching toward the fully stretched limit. (The orientation S tends to saturate at such large WiR and has virtually no effect on the hardening.) 3.4. Possible Origin of Lack of Hardening of Melts. As shown in the previous sections, the behavior of entangled solutions, i.e., thinning of ηE followed by thickening, is well captured by the

PCN simulation incorporating reptation, contour length fluctuations, thermal CR and CCR, and the FENE correction. Hence, the lack of hardening observed for entangled PS melts (Figure 3b) is quite possibly due to some extra factor other than those listed above. Some years ago an extra factor was proposed, related to the tube constraint and called interchain pressure (ICP),14 which generated the prediction ηE  ε·1/2 under steady, moderately fast, elongational flows. Later, tube models somehow incorporating ICP9 predicted monotonic thinning of ηE (ε·1/2) even in very fast flows, i.e., at WiR > 1, in agreement with the observed behavior of PS melts. However, combination of ICP and CCR, the latter being an essential ingredient in the tube model, results in excessive thinning stronger than the observation.16 In addition, there is no conceptual reason why the ICP effect should be active only in melts and not in entangled solutions where the constraining tube is also present. A possible candidate factor discriminating between entangled melts and entangled solutions is the anisotropic change of the friction coefficient tensor under strong flows. An anisotropic friction was proposed long ago by Giesekus,36 and further developed by Wiest.18 Their constitutive equations, incorporating the anisotropic friction for a dumbbell model, show that the steady elongational viscosity can decrease for WiR > 1. Gupta et al.21 showed that the constitutive equation by Wiest reasonably reproduces the extensional viscosity of dilute polymer solutions. However Wiest equation cannot be used for the case of entangled polymers because of the separate role played by Wi and WiR, where Wi controls orientation of the entangled subchains while WiR determines subchain stretch.1 For the case of entangled polymers, the flow-induced anisotropic reduction of the monomeric friction has recently been reconsidered by Ianniruberto and co-workers37 and later by Uneyama and co-workers38 in their rheo-dielectric analysis. Within the context of a flow-induced frictional anisotropy, the difference between entangled melts and solutions might arise from the fact that, in a monodisperse melt, each stretched and oriented subchain is surrounded by similarly stretched/oriented material, whereas in entangled solutions subchains are surrounded by solvent molecules that remain isotropic under all conditions. We are now attempting to evaluate the magnitude of the friction reduction from experimental data available in the literature so as to use the outcome in the PCN simulation without introducing any additional parameter. The results will be reported in a subsequent paper.

4. CONCLUDING REMARKS We have implemented finite extensibility (FENE) into the original PCN code to investigate the behavior of both transient and steady-state viscosity (ηE) under fast uniaxial elongation of entangled polymers. The PCN simulation with the FENE effect well mimicked the behavior of the entangled semidilute solutions, i.e., the thinning of ηE at extensional rates ε· between τd1 and τR1 then followed by nondivergent thickening at ε· > τR1. In the range τd1 < ε· < τR1, where finite extensibility is not yet important, the simulation revealed an important contribution of the CCR mechanism to the thinning behavior of the steady-state elongational viscosity, resulting in a decrease of the entanglement density with increasing ε·, and the corresponding reduction of the chain tension. Conversely, for ε· > τR1, the simulation shows that the thickening results from the enhancement of chain tension due to chain stretching toward the fully stretched limit. 9680

dx.doi.org/10.1021/ma202166y |Macromolecules 2011, 44, 9675–9682

Macromolecules

ARTICLE

These simulation features provide us with a clue for better understanding the elongational thinning/thickening behavior of entangled semidilute solutions. However, the PCN-FENE simulation cannot describe the behavior of entangled PS melts, i.e., monotonic thinning of ηE observed even at ε· > τR1. Thus, this monotonic thinning must be attributed to an extra mechanism not considered in the current PCN-FENE simulation. The stretch/orientation-dependent frictional anisotropy might be the factor that gives rise to monotonic thinning in fast extensional flows of entangled melts. Incorporation of the latter mechanism in the PCN-FENE simulation is now being attempted, and the results will be reported in a subsequent paper.

’ APPENDIX A. EFFECT OF SIMULATION BOX SIZE We have checked the effect of simulation box size to verify possible artifacts due to periodic boundary conditions. Figure 7 shows the time development of the (nondimensional) elongational viscosity for the longest chains examined in the paper (Z0 = 35), at WiR = 0.35 and WiR = 3.5, for various box dimensions. The simulation was always halted when the box dimension became equal to 4 in the directions normal to the uniaxial stretch. Figure 7a reports results obtained with several rectangular cuboid (flat) boxes having the same initial dimension in the stretching direction (set to 4) but different dimensions in the direction orthogonal to the stretch. In order to check for possible artifacts due to the use of flat initial boxes, we also performed simulations starting with cubic boxes of various sizes, and the corresponding results are shown in Figure 7b. Clearly, results are insensitive to both size and shape of the initial box. We used the largest box (with dimension 4  104  104) to examine the behavior in the steady flow state over a sufficiently long time. ’ APPENDIX B. DETERMINATION OF SIMULATION PARAMETERS The parameters for the melt were essentially common to our earlier studies. The subchain molecular weight M0 for PS melts has been reported earlier25 as 1.1  104. (Notice that the value of the molar mass between consecutive sliplinks is somewhat smaller than the entanglement molar mass Me reported in the literature.24,25) Using such a value, the subchain number Z0 was determined as Z0 = M/M0 (M = molar mass of the examined melts). The unit modulus G0 was evaluated as G0 = FRT/M0, with F and R being the density and gas constant, respectively. The unit time τ0 has been determined earlier25 as 2.0  103 s at 180 C, and the unit time at 130 C, τ0 = 5.6 s, was obtained from the WLF relationship reported in ref 25. The Rouse relaxation time τR was determined from eq 13. The newly √ introduced parameter in this study is the Kuhn length b = a/ n0 in eq 7 for the FENE effect. We determined n0 from the molar mass of the PS Kuhn segment reported as 720.4 As the value of n0 comes out relatively small (n0 = 15), a small FENE effect is present also in linear viscoelasticity. Hence a correction was necessary for the time constant τ0. The correction was made by fitting G*(ω), and we obtained τ0 = 7.0 s at 130 C. We determined the parameters for the solutions from the data reported in ref. 5. where the concentration dependence of the entanglement molecular weight Me was discussed in detail. Assuming that the ratio of the unit molecular weight M0 to Me is insensitive to the concentration (and thus the ratio is common with the melt; M0/Me = 1.10/1.33), M0 for the solutions were determined from Me at each concentration reported in ref 5.

Figure 7. Transient uniaxial viscosity for chains with Z0 = 35 at WiR = 3.5 and WiR = 0.35 obtained from simulations starting with various box dimensions. The simulations were started with (a) flat and (b) cubic boxes with the dimensions shown in the figure.

Figure 8. Linear viscoelastic G0 and G00 data of (a) PS solutions with M = 3.9  106 and c = 7.35 wt % (triangles) and c = 10 wt % (circles) from ref 5 and (b) PS melts with M = 2.0  105 (triangles) and M = 3.9  105 (circles) from ref 8. Simulation results are indicated with curves.

Then Z0, G0 and n0 were determined in a manner similar to that for melt. The unit time τ0 was obtained by fitting G*(ω) for each solution, and τR was calculated from Z0 and τ0. Since n0 values for the solutions are sufficiently large, the FENE correction for the time constant performed for the melt case was actually not necessary. 9681

dx.doi.org/10.1021/ma202166y |Macromolecules 2011, 44, 9675–9682

Macromolecules The parameters thus obtained are summarized in Table 1. The moduli simulated with those parameters (also accounting for FENE) are shown in Figure 8 (curves). For both solutions and melts (top and bottom panels), the simulated results agree with data (symbols) reasonably well, which lends support to the parameter values used in the simulation.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]. Present Addresses

ARTICLE

(30) Dambal, A.; Kushwaha, A.; Shaqfeh, E. Macromolecules 2009, 42, 7168. (31) Kushwaha, A.; Shaqfeh, E. J. Rheol. 2011, 55, 463. (32) Schieber, J. D.; Nair, D. M.; Kitkrailard, T. J. Rheol. 2007, 51, 1111. (33) Peterlin, A. J. Chem. Phys. 1960, 36, 1799. (34) Masubuchi, Y.; Uneyama, T.; Watanabe, H.; Ianniruberto, G.; Marrucci, G.; Greco, F. J. Chem. Phys. 2010, 133, 134902. (35) Uneyama, T.; Masubuchi, Y. J. Chem. Phys. 2011, 135, 184904. (36) Giesekus, H. J Non-Newtonian Fluid Mech. 1982, 11, 69. (37) Ianniruberto, G.; Brasiello, A.; Marrucci, G. Proc. 7th Annu. Eur. Rheol. Conf. 2011, 61. (38) Uneyama, T.; Horio, K.; Watanabe, H. Phys. Rev. E 2011, 83, 001800.

^

Fabricated Products Developments Division, Mitsui Chemicals Inc., Chiba 299-0265, Japan.

’ REFERENCES (1) Doi, M.; Edwards, S. The Theory of Polymer Dynamics, Oxford University Press: Oxford, U.K., 1986. (2) Watanabe, H. Prog. Polym. Sci. 1999, 24, 1253. (3) McLeish, T. C. B. Adv. Phys. 2002, 51, 1379. (4) Rubinstein, M.; Colby, R. Polymer Physics: Oxford: New York, 2003. (5) Bhattacharjee, P. K.; Oberhauser, J. P.; McKinley, G. H.; Leal, L. G.; Sridhar, T. Macromolecules 2002, 35, 10131. (6) Bhattacharjee, P. K.; Nguyen, D. A.; Mckinley, G. H.; Sridhar, T. J. Rheol. 2003, 47, 269. (7) Ye, X.; Sridhar, T. Macromolecules 2005, 38, 3442. (8) Bach, A.; Almdal, K.; Rasmussen, H. K.; Hassager, O. Macromolecules 2003, 36, 5174. (9) Wagner, M. H.; Kheirandish, S.; Hassager, O. J. Rheol. 2005, 49, 1317. (10) Luap, C.; M€uller, C.; Schweizer, T.; Venerus, D. C. Rheol. Acta 2005, 45, 83. (11) Mead, D. W.; Larson, R. G.; Doi, M. Macromolecules 1998, 31, 7895. (12) Ianniruberto, G.; Marrucci, G. J. Non-Newtonian Fluid Mech. 1996, 65, 241. (13) Marrucci, G. Polym. Sci. Polym. Phys. Ed. 1985, 23, 159. (14) Marrucci, G.; Ianniruberto, G. Macromolecules 2004, 37, 3934. (15) Marrucci, G.; Ianniruberto, G. J. Non-Newtonian Fluid Mech. 2005, 128, 42. (16) Dhole, S.; Leygue, A.; Bailly, C.; Keunings, R. J. Non-Newtonian Fluid Mech. 2009, 161, 10. (17) Jones, D. M.; Walters, K.; Williams, P. R. Rheol. Acta 1987, 26, 20. (18) Wiest, J. M. Rheol. Acta 1989, 28, 4. (19) Quinzani, L. M.; Armstrong, R. C.; Brown, R. A. J. Rheol. 1995, 39, 1201. (20) Li, X.; Denn, M. M. J. Rheol. 2004, 48, 805. (21) Gupta, R. K.; Nguyen, D. A.; Sridhar, T. Phys. Fluids 2000, 12, 1296. (22) Haward, S. J.; Sharma, V.; Odell, J. A. Soft Matter 2011, 7, 9908. (23) Masubuchi, Y.; Takimoto, J.-I.; Koyama, K.; Ianniruberto, G.; Greco, F.; Marrucci, G. J. Chem. Phys . 2001, 115, 4387. (24) Masubuchi, Y.; Ianniruberto, G.; Marrucci, G.; Greco, F. J. Chem. Phys. 2003, 119, 6925. (25) Masubuchi, Y.; Ianniruberto, G.; Greco, F.; Marrucci, G. J. NonNewtonian Fluid Mech. 2008, 149, 87. (26) Masubuchi, Y. Nihon Reoroji Gakkaishi (J. Soc. Rheol. Jpn.) 2006, 34, 275. (27) Yaoita, T.; Isaki, T.; Masubuchi, Y.; Watanabe, H.; Ianniruberto, G.; Greco, F.; Marrucci, G. J. Chem. Phys. 2008, 128, 154901. (28) Furuichi, H.; Nonomura, C.; Masubuchi, Y.; Watanabe, H. J. Chem. Phys. 2010, 133, 374902. (29) Graessley, W. W. Adv. Polym. Sci. 1982, 47, 67–117. 9682

dx.doi.org/10.1021/ma202166y |Macromolecules 2011, 44, 9675–9682